Psychology of Music - Jurusan Informatikafrdaus/PenelusuranInformasi/... · 4 Psychology of Music Despite a concerted effort since 2003 to measure personality in terms of the Big

http://pom.sagepub.com/Psychology of Music

http://pom.sagepub.com/content/early/2011/03/12/0305735610388897The online version of this article can be found at:

DOI: 10.1177/0305735610388897

published online 16 March 2011Psychology of MusicPeter Gregory Dunn, Boris de Ruyter and Don G. Bouwhuis

behavior, and personalityToward a better understanding of the relation between music preference, listening

Published by:

http://www.sagepublications.com

On behalf of:

Society for Education, Music and Psychology Research

can be found at:Psychology of MusicAdditional services and information for

http://pom.sagepub.com/cgi/alertsEmail Alerts:

http://pom.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

at SAN JOSE STATE UNIV on April 29, 2011pom.sagepub.comDownloaded from

http://pom.sagepub.com/

http://pom.sagepub.com/content/early/2011/03/12/0305735610388897

http://www.sagepublications.com

http://www.sempre.org.uk

http://pom.sagepub.com/cgi/alerts

http://pom.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav


Toward a better understanding of the relation between music preference, listening behavior, and personality

Peter Gregory DunnPhilips Research & University of Technology Eindhoven, the Netherlands

Boris de RuyterPhilips Research, the Netherlands

Don G. Bouwhuis University of Technology Eindhoven, the Netherlands

AbstractPrevious research relating personality and music preferences has often measured such reported preferences according to genre labels. To support previous research, the current paper has expanded investigation of the relation between personality and music preferences to include direct measurement of music listening behavior. A study (N = 395) measured participants’ personality, reported music preferences, and their listening behavior, which was tracked while using a music database for a minimum period of three months. Results indicated that reported music preferences were correlated to listening behavior, and indicated robust positive relations between Neuroticism and Classical music preference, and between Openness to Experience and Jazz music preference. Results also indicated issues when using genre labels to measure music preferences, which are discussed.

KeywordsBig Five, genre, listening behavior, music preferences, personal preferences, personality

Numerous factors are involved when an individual selects a particular song, album, or genre of music to be played (Levitin, 2006). Arguably, these factors include, but are not limited to: emo-tions, personal experience, social context, and culture. Nonetheless, Levitin has stated that per-sonality has a predictive influence over individuals’ music preferences. Levitin’s statement is

Psychology of Music1–18

© The Author(s) 2011Reprints and permission: sagepub.

co.uk/journalsPermissions.navDOI: 10.1177/0305735610388897

pom.sagepub.com

Corresponding author:Peter Gregory Dunn, Flow Interactive, 90–98 Goswell Road, London, EC1V 7DF, UK.Email: [email protected]

Article



2 Psychology of Music

mainly grounded in research that has related personality and music preferences (e.g., Arnett, 1992; Rentfrow & Gosling, 2003; Zweigenhaft, 2008). Further support, however, has come from research indicating that individuals see music preferences as strong indicators of person-ality characteristics (e.g., North & Hargreaves, 1999; Rentfrow & Gosling, 2006). North and Hargreaves have gone so far as to call music preferences a badge of identity, which individuals use as an indicator of their personality and help define the social groups that they attach them-selves to. Thus, there is a clear relation between personality and music preferences, but so far the vast majority of research that has investigated this relation has focused on using reported music genre preferences as its dependent variable. The current paper aims to support and build on previous research by incorporating individuals’ listening behavior to record music prefer-ences, which can then be related to both reported music preferences and personality.

Music preferences and personalityPrior to 2003, there had been several attempts made by researchers to develop their own meas-ure of music preferences or analyze the structure of music preferences (e.g., Cattell & Anderson, 1953; Litle & Zuckerman, 1986; Rawlings, Barrantes i Vidal, & Furnham, 2000; Rawlings, Hodge, Sherr, & Dempsey, 1995). Still, none of these studies had made a concerted effort to produce a comprehensive measure of music preferences based on rigorous testing of a music preference model over a series of studies. Motivated to create such a measure, Rentfrow and Gosling (2003) developed their own four-factor model of music preferences, which was then related to personality and other characteristics (e.g., self-views, cognitive ability). Since that landmark study, various researchers have tried to replicate Rentfrow and Gosling’s findings (e.g., Delsing, Ter Bogt, Engels, & Meeus, 2008; George, Stickle, Rachid, & Wopnford, 2007; Rentfrow & Gosling, 2006; Zweigenhaft, 2008). Consequently, Rentfrow and Gosling’s work represents a recent turning point in how music preferences according to genre have been meas-ured and their work has had a profound influence on related music preferences and personality research since 2003. For these reasons, the present paper has focused on Rentfrow and Gosling’s work and on related research since 2003.

Rentfrow and Gosling’s (2003) research and the research that has followed them (e.g., Delsing et al., 2008; George et al., 2007; Rentfrow & Gosling, 2006; Zweigenhaft, 2008) has been a substantial step forward toward understanding the structure of music preferences. Nonetheless, the research has had mixed results. For example, George and colleagues found an eight-factor model when they included 30 music genres, compared to the 14 genres included in Rentfrow and Gosling’s original model. Furthermore, both George and colleagues and Delsing and colleagues found subtle differences in the factor structure when comparing their model to Rentfrow and Gosling’s. On the one hand, rock, heavy metal, and alternative genres were con-sistently grouped together. On the other hand, genres like rap and dance/electronica, or blues, jazz, and classical, were inconsistently grouped; sometimes under the same factor and some-times not. Therefore, despite statements from these authors (i.e., Delsing et al., 2008; George et al., 2007) supporting Rentfrow and Gosling’s model of music preferences, their findings indi-cate that further research is needed.

The research investigating music preferences since 2003 (e.g., Delsing et al., 2008; George et al., 2007; Rentfrow & Gosling, 2003, 2006; Zweigenhaft, 2008) and much of the research prior to it (e.g., Arnett, 1992; Litle & Zuckerman, 1986; McNamara & Ballard, 1999) has almost exclusively relied on individuals’ self-reports to measure and broadly define music pref-erences according to genre. The subtle differences among factor structures describing music preferences suggest different notions of genre categorization among the different participant



Dunn et al. 3

samples. Consequently, it is argued that measuring reported music preferences by means of genre labels might be limited in its applicability. This argument is supported by Aucouturier and Pachet (2003), who contend genre categorization is inconsistent and indicate that there is no clear definition of what does, or does not, encapsulate a genre.

Given the argued genre limitations, it would seem prudent to better understand how reported music preferences are related to actual music listening behavior. Investigating the relation between reported music preferences and listening behavior might seem trivial, but it is well understood that attitudes are not always consistent with behavior. With respect to music pref-erences, research has clearly indicated the importance of music as a tool that individuals use to communicate their social identity (e.g., North & Hargreaves, 1999; Rentfrow & Gosling, 2006). Given that music is a badge of one’s social identity (North & Hargreaves, 1999), then it is rea-sonable to suggest that social desirability may at least subconsciously influence how one responds to questions regarding one’s music preferences. Therefore, though it seems likely that reported music preferences and listening behavior are strongly positively correlated, the differ-ing notions of genre among participants and the influence of social desirability when reporting music preferences warrants an investigation of the relation between these two variables.

Most researchers investigating the relation between personality and music preferences since Rentfrow and Gosling’s (2003) paper have measured personality in terms of the Big Five trait model rather than using various other measures based on different personality models (e.g., Chamorro-Premuzic & Furnham, 2007; Delsing et al., 2008; George et al., 2007; Rentfrow & Gosling, 2006; Zweigenhaft, 2008). Within personality trait theory, the Big Five was developed and robustly tested over a series of studies, and has become arguably the most accepted model of personality traits (John & Srivastava, 1999). As its name implies, the Big Five measures five trait dimensions (Costa & McCrae, 1992), which have been identified and described as:

Neuroticism (N): an individual’s propensity to feel fear, sadness, embarrassment, anger, guilt, and other emotions of negative affect.

Extraversion (E): an individual’s propensity to be sociable, talkative, assertive, active, and indicates their preference toward stimulating and exciting environments.

Openness to experience (O): an individual’s propensity toward intellectual curiosity, imagination, aesthetic and emotional sensitivity, and originality.

Agreeableness (A): an individual’s propensity toward being altruistic, helpful, sympa-thetic, and empathetic toward others.

Conscientiousness (C): an individual’s propensity toward cleanliness, orderliness, having self-determination, and self-control.

Each of these dimensions represents a continuous scale with opposite extremes. Higher scores for a given dimension are interpreted such that the individual should be more consistent in personality with the dimension label and description (e.g., extraversion), whereas lower scores are interpreted such that the individual should be more consistent with personality adjectives that are opposite to the dimension label and description (e.g., introversion). Furthermore, Costa and McCrae (1992) have developed a measure of the Big Five that also assesses trait facets. There are six facet trait descriptors within each of the dimensions, which are measured and interpreted in the same manner as their affiliated trait dimensions. For instance, individuals who score high on the aesthetics facet under openness to experience are more open-minded about (new) aesthetic experiences, while the opposite can be said of individuals who score low on this facet. As a result, facets offer more specific and detailed personality information.




Despite a concerted effort since 2003 to measure personality in terms of the Big Five, the research correlating music preferences with the Big Five has provided mixed results as well. These mixed correlation results are demonstrated in Table 1, which summarizes previous cor-relations found between music preferences and the Big Five. The first column provides the origi-nal four music preference dimensions presented in Rentfrow and Gosling’s four-factor model, followed by the genres contained within each of these dimensions in the second column. The third through sixth column indicate the significant correlations between music preferences by genre and abbreviated traits for each of the referenced research papers shown as column head-ings: (1) Rentfrow and Gosling (R & G; 2003); (2) Delsing et al. (D et al.; 2008); (3) George et al. (G et al.); and (4) Zweigenhaft (Z; 2008). While there are a number of consistent findings among the studies summarized in this table, it is evident that there are also several inconsisten-cies across the studies as well. Indeed, there are conflicting results (e.g., pop, rap/hip-hop) in which some research has reported a positive correlation for a given trait, while other research has reported a negative correlation for the same trait.

One possible explanation for the inconsistencies summarized in Table 1 is again attributable to the genre ambiguity issue described earlier. That is, differing notions of music genres among different populations are likely to reveal different relations between personality and music pref-erences. Investigating the relation between personality and both reported music preferences and listening behavior might help identify more robust relations between these variables. Furthermore, most of the studies referred to in Table 1 measured participants’ personality at the dimension level. It has been argued that facets would provide greater descriptive detail and a better understanding of the relation between personality and music preferences (Rentfrow & Gosling, 2003; Zweigenhaft, 2008). While Zweigenhaft (2008) did correlate reported music

Table 1. Significant correlations found between the Big Five dimensions and music preferences in research studies since 2003

Music Dimension Genre Correlated Traits

R & G D et al. G et al. Z

Reflective & complex Blues O – O OClassical O O, N O -Folk O – E, C OJazz O O, N O O

Intense & rebellious Alternative O – O, A, C -Heavy metal O O O, A, C -Rock O O O, A, C -

Upbeat & conventional Country E, A, C, O – E, C -Pop E, A, C, O E, A O, A, C OReligious E, A, C, O – - OSoundtracks E, A, C, O – – A, O

Energetic & rhythmic Dance/ Electronica E, A E, A O, C -Rap/hip-hop E, A E, A O, A, C E, OSoul/funk E, A E, A – O

Notes: Referenced material: R & G Rentfrow & Gosling, 2003; D et al. Delsing et al., 2008; G et al. = George et al., 2007; Z Zweigenhaft, 2008. Trait abbreviations: N Neuroticism; E = Extraversion; O Openness; A Agreeableness; C Conscientiousness. Abbreviations denote significant correlations (p < .05) between trait and genre. Correlation is positive unless an underlined abbreviation is shown, indicating a negative correlation. Single dashes (-) indicate no significant correlations found in that particular study. Double dashes (–) indicate that the genre was not considered in that particular study.



Dunn et al. 5

preferences to personality facets, he also reported over 200 correlations doing so, which made it highly likely that spurious correlations were also reported. Reporting only significant correla-tions that are consistently found between personality and both reported music preferences and listening behavior would minimize the chance of spurious correlations being reported.

In sum, the present paper contributes to the research that has investigated the relation between music preferences and personality by including an observed measure of music listen-ing behavior. In doing so, the present paper aims to work toward a better understanding of the structure of music preferences, clarify the relation between reported music preferences and listening behavior, and build on and give greater detailed insight into the patterns of music preferences and its relation to personality. Based on the reviewed literature, the hypotheses for this study were as follows:

Music preferences data from the current sample will support Rentfrow and Gosling’s (2003) model of music preferences.Reported music preferences will be positively correlated with listening behavior for the same genre.Significant correlations between reported music preferences and personality will be con-sistent with significant correlations obtained between music listening behavior and per-sonality for the same genres.

Method

ParticipantsParticipants (N = 395; 335 males) volunteered following a recruitment announcement adver-tised to individuals using an experimental music database (see materials). All participants were employees of Royal Philips Electronics. Participants’ ages in this sample ranged from 22 to 60 years (M = 36.7, SD = 8.93). Five participants did not provide their age. There were 29 nation-alities represented in this sample. Most participants were Dutch (n = 202), but other nationali-ties included the United States (US; n = 50), France (n = 35), Germany (n = 18), Belgium (n = 16), United Kingdom (UK; n = 11), Other European countries (n = 33), Latin Americas (n = 4), Canada (n = 2), and Asia/Pacific (n = 10). Fourteen participants did not specify their national-ity. Due to attrition, not all parts of the study were completed; 267 participants (227 males) provided sufficient listening behavior data (see procedure), and of those participants, 138 (114 males) completed the personality measure (NEO PI-R). Participants’ mean age for the first sub-sample (n = 267) was M = 36.5 years (SD = 8.77). Mean age for the second sub-sample (n = 138) was M = 36.4 years (SD = 8.71). Nationalities for these sub-samples were proportionally similar to the complete sample.

MaterialsThe music database used was an experimental platform available to participants via Royal Philip Electronic’s intranet. This database contained nearly 70,000 audio recordings which were originally uploaded by its users. These recordings were tagged according to an industry standard (All Music Guide [AMG], 2008) into one of 16 music genre categories: alternative, blues, classical, country, dance, folk, funk, heavy metal, jazz, pop, rap, religious, rock, R&B, soundtracks, and ‘other’ category. The other category included miscellaneous items (e.g.,




underground music, comedy). With exception to the R&B and other categories, these genres matched the 14 genres used by Rentfrow and Gosling (2003). Participants’ music listening behavior was measured in two ways:

Song count tracked the number of songs selected for listening, per genre, by each partici-pant. For each participant, the number of songs selected was divided by the total number of songs they listened to. Thus, this dependent variable was the relative percentage of songs that began playing within each genre for each participant.1

Listening duration tracked the time duration (in seconds) of music listened to, per genre, by each participant. For each participant, their time duration per genre was divided by their total listening time. Thus, this dependent variable was the relative listening time percentage within each genre for each participant.

A minimum criterion of at least 100 songs selected for listening was imposed to estimate par-ticipants’ typical listening behavior. This meant that participants’ minimum amount of time listening to music was roughly 200 minutes. In addition to tracking participants’ music listen-ing behavior, two psychometric measures were used in the study:

Short Test of Music Preference (STOMP) measured participants’ reported music prefer-ences (Rentfrow & Gosling, 2003). Participants rated their music preference toward 14 genre items. These items load onto the four dimensions described in Rentfrow and Gosling’s model of music preferences. Items were rated on a scale from 1 (strongly dislike) to 7 (strongly like).Revised NEO Personality Inventory (NEO PI-R) measured participants’ personality (Costa & McCrae, 1992). Participants rated 240 items on a scale from 1 (strongly disagree) to 5 (strongly agree), which load onto the Big Five personality dimensions. The NEO provided aggregated scores for the five personality dimensions, as well as scores for the six facets contained within each dimension. Participants were able to complete the NEO PI-R in either English (Costa & McCrae, 1992), or in Dutch (Hoekstra, Ormel, & de Fruyt, 2003).

ProcedureAfter providing informed consent, participants were given the option to complete a survey in either English or Dutch. The survey consisted of some demographic information (age, gender, nationality, and years of musical training), the STOMP, and the NEO PI-R. Once the entire sur-vey had been completed, participants were debriefed and thanked for their participation. If the participant had completed the NEO PI-R, they were also provided with a personality report as reward. The music database was only accessible to participants while at their office desk.

The music database recorded participants’ music listening behavior prior to them receiving the survey and then further tracked their listening behavior subsequent to completing the sur-vey for a minimum period of three months. It is possible that participants’ listening behavior may have been influenced by knowing the purpose of the study provided in the debriefing. Nonetheless, most of the collected listening behavior data was taken prior to participants’ com-pletion of the survey. Furthermore, it was believed that participants’ listening behavior would not be unusually influenced for the entire three-month tracking period given the information provided in the debriefing.



Dunn et al. 7

Results

Confirming the existing model of music preferencesA large enough sample (N = 395) had been collected to test the first hypothesis and conduct a Confirmatory Factor Analysis (CFA) of the music preference dimensions specified by Rentfrow and Gosling (2003). Using LISREL (Jöreskog & Sörbom, 2007), the CFA was conducted with participants’ music preference ratings taken via the STOMP. In addition to a chi-square ( 2), several other goodness-of-fit criteria were used to assess the adequacy of the model (Tabachnick & Fidell, 2007). The criteria included the Comparative Fit Index (CFI), the Root Mean Square Error of Approximation (RMSEA) and the Standardized Mean Square Residual (SRMR). A CFI above .95 indicates a good-fitting model, while decision rules regarding the cut-off criteria for RMSEA and SRMR indicate that values should be below .10 and .08, respectively (Tabachnick & Fidell, 2007). The Goodness of Fit Index (GFI) and the Adjusted Goodness of Fit Index (AGFI) were also calculated to provide estimates of the proportion of variance accounted for by the model. Figure 1 illustrates the standardized parameter estimates for the CFA model given the obtained music preference data.

A significant chi-square was calculated, 2 (71, N = 395) = 499.27, p < .001, while other fit criteria statistics were: CFI = .56, RMSEA = .12, SRMR = .10, GFI = .77, AGFI = .85. Taken together, the present results suggested that, unlike Rentfrow and Gosling’s (2003) results, the obtained data did not fit the existing model well.

Figure 1. Standardized parameter estimates for CFA model from the obtained music preference data. Notes: 2 (71, N = 395) 499.27, p < .001 (GFI .77, AGFI .85, CFI .56, RMSEA .12, SRMR .10). e error variance.




The existing model was based on results from a series of rigorous studies conducted by Rentfrow and Gosling (2003). So, it was important to better understand how and why obtained data did not fit the existing model. Visual examination of the CFA results provided in Figure 1 suggests that pop provided the only noticeably weak path coefficient. Conceivably then, remov-ing pop would provide a much better fit for the CFA model given the obtained data. Though removing pop from the model did improve the fit criteria statistics, further removal of sound-tracks and jazz was necessary before the model showed minimally acceptable fit criteria statis-tics. After removal of these three genre variables, the fit criteria statistics were, 2 (38, N = 395) = 173.59, p < .001, CFI = .79, RMSEA = .095, SRMR = .071, GFI = .93, AGFI = .87. Alternatively, additional paths between independent genre variables and latent music prefer-ence dimensions could be added until a better model fit using the obtained data is reached. Unfortunately, results showed that numerous paths would have to be added and still a good model fit would not be reached. Overall, these results suggest that the patterns of music prefer-ences toward genres with the obtained data differ substantially from the existing model pro-vided by Rentfrow and Gosling (2003).

To further investigate how and why obtained data differed from Rentfrow and Gosling’s (2003) data and model, Principal Components Analysis (PCA) used obtained STOMP ratings to explore alternative music preference dimensions. The PCA was done using SPSS 15.0 (SPSS, 2006) and used Varimax rotation. Results from the PCA were somewhat conflicted. Initially, a six-factor solution that accounted for 70% of the total variance was obtained by implementing the Kaiser rule (eigenvalues greater than 1), which were supported by results from a parallel analysis with Monte Carlo data (O’Connor, 2000). Table 2 provides the six-factor, Varimax-rotated PCA solution. Cells in Table 2 indicate the factor loading for the indicated genre (rows) and factor (columns). Factor loadings marked in bold indicate the highest loading, signifying that the predicted variance in that genre is explained most by that factor. With exception to the Bass-Heavy label, these factors were labeled based on genre categorization by AMG (2008). The

Table 2. PCA factor loadings from the 14 genres using a six-factor, varimax-rotated solution

Genre Music preference dimension

Rhythm ‘n’ blues Hard rock Bass heavy Country Soft rock Classical

Jazz .774 .069 .097 .154 .017 .278Blues .754 .006 .182 .311 .001 .061Soul .703 .063 .383 .113 .072 .143Heavy Metal .061 .812 .083 .134 .024 .141Alternative .134 .763 .161 .077 .143 .222Rock .106 .655 .176 .057 .548 .110Rap .113 .017 .842 .119 .056 .089Dance .010 .111 .763 .109 .056 .098Country .020 .112 .007 .834 .145 .069Folk .146 .164 .016 .731 .079 .118Pop .077 .002 .097 .015 .869 .155Soundtracks .157 .061 .149 .079 .613 .507Classical .222 .013 .132 .043 .020 .762Religious .024 .016 .163 .429 .140 .603

Notes: N 395. All factor loadings .300 or larger are provided in italics; factor loadings in bold represent highest factor loadings for each genre given each dimension.



Dunn et al. 9

Bass-Heavy label was used to express an audio characteristic generally perceived to be exhibited in that music.

Unfortunately, the initial six-factor solution conflicted with results using Goldberg’s (2006) Bass Ackwards method to determine how many principle components should be retained. Instead, the Bass Ackwards method suggested a four-factor solution that accounted for 52% of the total variance. Nevertheless, it should be said that this four-factor solution was markedly different from Rentfrow and Gosling’s (2003) four-factor model of music pref-erences. Table 3 gives the structure of the four-factor Bass Ackwards solution and is organ-ized in the same manner as Table 2. Table 3 also provides the opportunity to compare its structure of music preferences to that of Rentfrow and Gosling’s model outlined in Table 1. The four factors were named in similar fashion to the six-factor solution. Indeed, there are two factors that have identical compositions and so have the same factor name. The remain-ing two factors from the four-factor solution have been named to express core audio quali-ties of the type of music represented by each of these factors. Even here there is a clear connection between the Bass Foundation factor in the four-factor solution and the Bass-Heavy factor in the six-factor solution. Connections between the four-factor and six-factor solutions are further facilitated by Figure 2, which provides a hierarchical representation of a one-through-six-factor solution using the Bass Ackwards method. In Figure 2, correlation coefficients greater than .40 are shown between factors at each level to show the progres-sion from a one-factor solution to a six-factor solution. In doing so, Figure 2 summarizes how and why the data obtained from the current study substantially differed from the data used by Rentfrow and Gosling (2003) to construct their model of music preferences. Given the obtained results, it seemed prudent to conduct further analyses at the genre level, rather than using Rentfrow and Gosling’s music dimensions.

Table 3. PCA factor loadings from the 14 genres using a four-factor, varimax-rotated solution

Genre Music preference dimension

Acoustic foundation Hard rock Bass foundation Soft rock

Folk .639 .108 .023 .101Country .624 .116 .051 .382Religious .606 .176 .082 .118Blues .600 .187 .146 .299Classical .568 .006 .030 .086Rock .060 .814 .077 .286Heavy metal .038 .723 .027 .068Alternative .096 .643 .190 .180Rap .036 .057 .792 .206Dance .127 .005 .693 .174Soul .245 .211 .649 .175Jazz .367 .220 .441 .398Pop .083 .261 .174 .657Soundtracks .232 .007 .089 .650

Notes: N 395. All factor loadings .300 or larger are provided in italics; factor loadings in bold represent highest factor loadings for each genre given each dimension.




Reported music preferences versus listening behaviorTo test the second hypothesis, analysis at the genre level compared participants’ reported music preferences to their listening behavior. Due to insufficient listening behavior from some of the participants, this analysis used the first sub-sample (n = 267) reported in the method section. Furthermore, correlations between the two measures of listening behavior (i.e., song count and listening duration) showed these measures to be largely redundant. No correlation was greater than r = .97 between these measures for each of the 16 genre categories tested. Therefore, only listening duration percentages were used for the remainder of the analyses because this measure was arguably slightly more accurate as a measure of participants’ entire music listening behavior (e.g., participants may not listen to an entire song after selecting it, and classical songs tend to be longer than songs from other genres). For convenience, listening duration percentages will be referred to as duration scores from this point on.

Prior to testing the second hypothesis, a 2 (language) ! 2 (gender) ! 16 (genre) mixed analy-sis of variance (ANOVA) was conducted to find out if participants’ duration scores differed depending on language or gender for the 16 genres tracked. There were no between-group effects due to language F(1, 258) = 0, ns, or gender, F(1, 258) = 0, ns. Additional multiple lin-ear regressions were conducted separately for musical training and age given duration scores across all genres to test for these effects. No effects were found for musical training on Duration scores, R2 = .05, F(15, 196)2 = .75, ns. There was an effect of age on Duration scores, however, R2 = .17, F(15, 247) = 3.48, p < .001, which indicated that age significantly predicted folk, partial = .16, t(250) = 2.83, p < .01, and pop duration scores, partial = .26, t(250) = 3.68, p < .001. Given that age accounted for a significant proportion of variance in only two of 16 genres, it was decided that it was not necessary to use age as a covariate for music preferences in further analyses. Therefore, there was no need to compare results separately for gender or language, or account for musical training or age in further analyses.

Figure 2. Bass Ackwards (Goldberg, 2006) hierarchical representation of a one-through-six-factor solution that indicates correlations between factors from one level to the next. Only correlations with a magnitude greater than .40 are shown



Dunn et al. 11

At this point, it is worth mentioning that the distribution of music content in the experimen-tal database was not evenly divided across the 16 genres used. Figure 3 provides a histogram indicating the relative percentages of songs available in the music database per tracked genre. The songs available in the music database had been added to this database by its users, which included participants from the current study. Consequently, the histogram describes the music preferences among all individuals using the music database. Interestingly, these music prefer-ences reflect the current state of industry music sales in the UK and US (see British Phonographic Industry [BPI], 2008; Information Technology (IT) Facts, 2008, respectively), particularly with respect to rock and pop genres. Furthermore, a correlation across genres between the percent-age of songs available on the database and the mean duration scores was r = .99. The magni-tude of this correlation might suggest that participants’ listening behavior is equal to chance probabilities solely dependent on the amount of music available for a given genre. Nonetheless, if the second hypothesis is confirmed, it will indicate that individuals will seek out the music that they enjoy regardless of the music that is available to them.

Ultimately, the second hypothesis was tested by correlating participants’ reported music preferences, via the STOMP, and their listening behavior, by means of their duration scores. Correlations found between participants’ STOMP ratings and their duration scores across gen-res are given in Table 4. As seen along the diagonal in Table 4, participants’ STOMP ratings were nearly always significantly positively correlated to their duration scores for the same genre. The lone exception to this trend was for alternative. Interestingly, however, Table 4 also shows that both combinations of preference ratings and duration scores for alternative and Rock were significantly positively correlated. Alternative music is often considered a sub-genre of rock (AMG, 2008). Furthermore, while no R&B rating was recorded by the STOMP, R&B can be considered synonymous with soul (AMG, 2008). Given this, results from Table 4 show that funk/soul ratings were significantly positively correlated with R&B duration scores.

Figure 3. Histogram indicating the relative percentages of songs per tracked genre compared to all available songs in the music database.




Tabl

e 4.

Cor

rela

tion

coef

ficie

nts

betw

een

part

icip

ants

’ STO

MP

scor

es a

nd th

eir

List

enin

g D

urat

ion

scor

es p

er g

enre

Mus

ic

List

enin

g

dura

tion

genr

e

STO

MP

Gen

re

Clas

sica

l Bl

ues

Jazz

Fo

lk

Alte

rnat

ive

Rock

H

eavy

M

etal

Co

untr

y Po

p R

elig

ious

So

undt

rack

s R

ap

Soul

D

ance

Clas

sica

l .3

3

.10

.0

2.0

0

.11

.1

1

.21

.0

2

.04

.0

6

.03

.0

8

.10

.10

Bl

ues

.00

.2

2

.05

.0

8

.00

.0

2

.03

.0

4

.06

.0

5

.14

.0

1

.16

.1

2

Jazz

.0

9.0

6.3

7

.00

.0

6

.00

.0

4

.15

.0

5

.05

.2

5

.02

.1

4*

.05

Fo

lk

.08

.0

1

.05

.1

6

.03

.0

6

.01

.1

1

.03

.0

4

.03

.0

5

.02

.0

1

Alte

rnat

ive

.11

.13

.0

4

.06*

.1

1.1

3

.16

.0

3

.05

.06

.0

1.0

6.0

5.1

3

Rock

.2

1.0

0

.01

.0

5.3

8

.33

.4

4

.13

.0

0

.12

.0

9

.00

.0

4

.07

H

eavy

met

al

.17

.15

.1

1

.10

.1

1

.11

.2

8

.11

.0

6

.10

.09

.0

0

.04

.0

5

Coun

try

.01

.0

4

.05

.1

8

.10

.0

9

.04

.2

7

.04

.0

7

.03

.0

1*

.06

.0

3

Pop

.10

.10

.1

2

.00

.3

3*.0

4

.13

.2

2

.20

.0

3

.07

.1

4

.08

.2

4

Rel

igio

us

.06

.0

0

.02

.0

6*

.08*

.2

9

.12

.0

5

.18

.2

6

.02

.0

7

.08

.1

0

Soun

dtra

cks

.05

.0

3.0

9

.05

.0

6*

.03

.1

3

.12

.0

6

.04

.1

4

.00

.1

2

.00

R

ap/h

ip-h

op

.10

.0

4

.01

.1

1

.04*

.0

3

.02

.0

1

.08

.08

.0

7

.42

.1

7

.15

So

ul /f

unk

.14

.2

3

.02

.12

.1

0*

.05

.0

1

.07

.0

5

.10

.1

5

.04

.24

.0

7

Dan

ce/

Elec

tron

ica

.03

.0

8

.04

.1

2.0

2

.17

.0

7

.16

.0

9

.01

.0

4

.13

.0

5

.43

R&

B .0

2

.07

.1

4

.13*

.1

6*

.13

.1

3

.03

.0

5

.08*

.0

6

.12

.1

6.0

7

Oth

er

.18

.0

1

.00

.1

0

.16*

.1

4

.29

.1

3

.05*

.0

8

.21*

* .0

7

.13*

.03

Not

e: N

2

67. C

orre

latio

n va

lues

in b

old

indi

cate

exp

ecte

d si

gnifi

cant

pos

itive

cor

rela

tions

bet

wee

n pa

rtic

ipan

ts’ r

epor

ted

mus

ic p

refe

renc

es a

nd th

eir l

iste

ning

beh

avio

r fo

r the

sam

e ge

nre.

* p

.05;

** p

.0

1.



Dunn et al. 13

Personality, music preferences, and listening behaviorThe third hypothesis was tested by correlating participants’ measured personality traits with both their reported music preferences and their listening behavior. Due to incomplete or unre-turned NEO PI-R surveys, these analyses used the second sub-sample (n = 138) reported in the method section. Table 5 provides correlations between participants’ measured personality dimensions and reported music preferences/listening behavior per genre. Columns separate correlations by each of the Big Five personality dimensions, further divided by correlations between participants’ measured personality dimension and their reported music preferences (S), or their duration scores (D). Rows separate correlations by genre. Looking at Table 5, only two pairs of correlations provided consistent significant findings between participants’ person-ality and their music preferences. These consistent correlations were between neuroticism and classical and between openness to experience and jazz.

In addition to investigating the relations between reported music preference or listening behavior and personality at the Big Five dimension level, the inventory used to measure person-ality afforded the opportunity to investigate the same relations at the facet level. There were 870 correlations calculated between reported music preference/listening behavior and person-ality. Of the 870 correlations, 79 were significant and there was a certainty that some of these correlations were spurious. So, to avoid reporting so many possibly spurious correlations, a criterion was specified so that only pairs of significant correlations between personality and both reported music preferences and listening behavior are reported. Using this criterion, only four pairs of significant correlations were found: (1) self-consciousness (neuroticism facet) was correlated with classical preference ratings (r = .27, p < .01) and with classical duration scores (r = .24, p < .01); (2) aesthetics (openness facet) was correlated with jazz preference ratings (r = .21, p < .05) and with jazz duration scores (r = .27, p < .01); (3) ideas (openness facet) was

Table 5. Correlations between participants’ personality traits and their reported music preferences/listening behavior per genre

Genre N E O A C

S D S D S D S D S D

Blues ".13 ".15 .05 .06 ".04 ".06 .09 .06 ".03 .00Classical .20* .18* ".08 ".04 .15 .14 .02 .02 ".09 ".24**Folk .06 .26** ".03 ".18* .10 .11 .20* .05 .00 ".05Jazz .08 .11 .04 ".15 .27** .18* ".05 .01 ".19* ".15Alternative .11 ".06 .05 .03 .18* .06 ".04 .11 ".09 .03Heavy metal ".04 ".02 .10 .03 ".04 .02 ".18* ".11 .05 .09Rock ".04 ".15 .02 .11 .20* .02 .07 ".09 .06 .08Country .05 .14 .01 .10 ".16 .00 .13 .15 ".10 .09Pop ".16 ".13 .20* .03 .02 ".10 .23** .09 .13 .16Religious .11 ".06 ".01 .17* .01 .05 .04 .09 ".13 .04Soundtracks .15 .04 .10 ".14 ".04 .09 .01 .18* .07 .05Dance/eltrnc ".15 .03 .22* ".07 .02 ".11 .06 .05 .09 .01Rap/hip-hop ".18* ".07 .21* .05 .04 ".04 .00 ".07 .14 .00Soul/funk .05 .02 .03 ".03 .13 ".17* .00 ".03 ".05 ".11

Notes: N 138. S STOMP preference ratings, D Duration scores. Items in bold represent agreement of significant correlations regardless of measure.* p .05; ** p .01.




correlated with jazz preference ratings (r = .23, p < .01) and with jazz duration scores (r = .21, p < .05); and (4) excitement-seeking (extraversion facet) was correlated with rock preference ratings (r = .19, p < .05) and with rock duration scores (r = .23, p < .01).

DiscussionThe present study focused on supporting previous research that has investigated the relation between personality and reported music preferences by expanding on that research to include a measure of music listening behavior. As expected, participants’ reported music preferences toward various genres were nearly always positively correlated with their listening behavior toward the same genre; 16 of the 17 correlations were significant. The study also attempted to confirm Rentfrow and Gosling’s (2003) model of music preferences, and correlations between music preferences and personality. The results are discussed in the following sub-sections in order of the hypotheses, beginning with the attempt to confirm Rentfrow and Gosling’s model of music preferences.

Confirming the existing model of music preferencesA CFA was employed to test the first hypothesis, which was concerned with whether or not Rentfrow and Gosling’s (2003) model of music preferences would be confirmed. Given the obtained results, the structure of Rentfrow and Gosling’s model was not confirmed. Additional analyses were subsequently done to explore structural differences in music preferences observed in the current sample compared to Rentfrow and Gosling’s sample. The additional analyses included a Principal Components Analysis (PCA), which initially revealed a six-factor structure using the current sample versus Rentfrow and Gosling’s four-factor structure. While parallel analysis supported the six-factor structure, Goldberg’s Bass Ackwards approach yielded a four-factor structure that was noticeably different to Rentfrow and Gosling’s four-factor structure.

The differing factor structures demonstrate subtle inconsistencies across the results pre-sented by various researchers. For instance, while rap and dance genres grouped together in the present research, these genres are grouped separately in other research (e.g., Delsing et al., 2008; George et al., 2007). Conversely, blues and jazz were grouped separately from classical in the six-factor PCA, but grouped together in other research (e.g., Delsing et al., 2008; Rentfrow & Gosling, 2003). Ultimately, only alternative, rock, and heavy metal genres were consistently grouped across all factor structures obtained from the present research and in other research (e.g., Delsing et al., 2008; George et al., 2007). While Rentfrow and Gosling labeled this factor as intense and rebellious, it was labeled hard rock in the current paper in agreement with other researchers (i.e., Delsing et al., 2008; Zweigenhaft, 2008), who have questioned the original label used by Rentfrow and Gosling. Consequently, the hard rock label suggests that this factor is more a reflection of the industry nomenclature of rock music (AMG, 2008), rather than thematic attitudes conveyed by Rentfrow and Gosling’s labels.

In sum, the current results suggest that there are differences among samples and cultures regarding how genre labels are viewed; what content is represented by a genre label, and how it is related to other genre labels. The distinctly different sample obtained in the present research (i.e., working adults) compared with much of the previous research (i.e., university students) suggests that age differences also influence how some genre labels are viewed and inevitably listened to. Age-related differences were certainly evident when considering the amount that participants listened to folk and pop music. Specifically, as participants’ age increased, so did the



Dunn et al. 15

amount that they listened to both pop and folk music. Whether these age-related differences are due to changes in people’s perception of genre over time or due to the evolving nature of the genre content cannot be answered by the present results. Nevertheless, analysis of the current results does benefit from introduction of the Bass Ackwards method, which helps to provide some insight into how various genre labels are viewed and the extent that they related to each other. In conclusion, the current results indicate that the ambiguous nature of genre labels (Aucouturier & Pachet, 2003) makes it difficult to construct a lasting and robust model of music preferences based on such genres.

Reported music preferences versus listening behaviorAs previously stated, participants’ reported music preferences were generally correlated to their listening behavior. Given the distribution of music content provided in the experimental data-base, the results show that participants specifically sought out and listened to music from gen-res they reportedly preferred. Analyses of these results also indicated no gender effects on the amount of music listened to for any particular genre, but it should be mentioned that the small number of females in the current sample suggests that gender effects on music listening behav-ior needs further investigation.

The results from this analysis also further support the notion that it is somewhat difficult to accurately measure music preferences using ambiguous genre labels. Specifically, the percep-tion of what a genre label represents is often confused and overlapping, which was evidenced by significant correlations between rock and alternative genres for example. The genre ambiguity argument might be particularly true for genre labels that are broadly defined (e.g., pop), poten-tially vaguely conceived by our participants (e.g., folk), or both (e.g., soundtracks). Therefore, it is argued that genre ambiguity contributed to lower correlations found amongst the results by introducing measurement error.

Personality, music preferences, and listening behaviorGenre ambiguity might also partly explain why correlation results between music preferences and the Big Five personality traits have varied so greatly across the research in this area (e.g., Delsing et al., 2008; George et al., 2007; Rentfrow & Gosling, 2003; Zweigenhaft, 2008). Such inconsistencies were also found in the current results. On the one hand, there were several sig-nificant correlations when considering reported music preferences, such as positive correla-tions between extraversion and pop, dance, and rap genres, which matched some of the previous research (e.g., Delsing et al., 2008; Rentfrow & Gosling, 2003). On the other hand, listening behavior data provided positive correlations between extraversion and religious music, as well as between agreeableness and Soundtracks; which is again similar to previous findings (e.g., Rentfrow & Gosling, 2003). Nonetheless, the only consistent correlations found after consider-ing both reported music preferences and listening behavior was between neuroticism and clas-sical music, as well as between openness to experience and jazz. Furthermore, only the latter of these two correlations was also consistent when compared with previous research (e.g., Delsing et al., 2008; George et al., 2007; Rentfrow & Gosling, 2003; Zweigenhaft, 2008).

Given the number of correlations that were performed, the results suggest that relatively few reliable correlations exist between personality and music preferences. Moreover, similar to previous research (e.g., Delsing et al., 2008; George et al., 2007; Rentfrow & Gosling, 2003; Zweigenhaft, 2008), most correlations between personality and music preferences are relatively




weak (i.e., r < .30). Consequently, if more robust and reliable relations between personality and music preferences are to be found, then measurement of both personality and music preferences must be more accurate. For example, preference for objectively measured music characteristics might provide greater accuracy for measuring music preferences when compared to reported genre preference.

It is noteworthy to point out that the consistent correlations found between personality and music preferences were improved when moving from personality dimensions to more specific personality facets. It is argued that the improvement is due to emphasizing the specific facets within a personality dimension that exhibit a stronger relation with a given music genre prefer-ence, while dismissing weaker relations exhibited by other facets within the same personality dimension with the same genre. In sum, these results suggest that including a measure of music listening behavior when examining the relation between music preference and personal-ity will limit spurious and inconsistent research findings and will help emphasize more robust and long-lasting findings.

Limitations and conclusionsThe results provided in the current article demonstrate that reported music preferences reflect music listening behavior. Nonetheless, more research investigating music listening behavior and its relation to personality would prove useful. For instance, while it is likely impossible to study participants’ music listening behavior in all contexts, it is clear that the present study has provided a limited context in this respect. Given this limitation, future research should focus on investigating music listening behavior in other non-occupational contexts and relating this to music preferences. Furthermore, the limitations of genre argued in the current paper suggests that research relating personality and music preferences would be served well if robust connec-tions between personality and specific, objective audio characteristics within various pieces of music were identified. Finally, the improved correlations observed using personality facets com-pared to the Big Five dimensions suggests that using these facets in future research could pro-vide a more specific and descriptive relation between personality and music preferences.

In conclusion, the current paper indicates that reported music preferences are generally related to listening behavior to the same genres. Nevertheless, it has also been argued that using genre to categorize music preferences presents its own limitations, as evidenced by the results. Specifically, the evolving and ambiguous nature of genre representation makes it diffi-cult to find long-lasting relations between music preferences and personality that provide robust results across cultures. Therefore, the inclusion of both measures of reported music preferences and listening behavior has brought us one step closer to a better understanding of the robust relations between music preferences and personality. In turn, this provides greater knowledge about the social relationships formed between individuals with similar tastes in music, and similarities in personality. Music is an integral part of individuals’ daily lives (Levitin, 2006), and so, unraveling the relation between personality and music preferences will inevita-bly provide practical and relevant information about what makes us who we are.

AcknowledgmentsThe work described in this paper was funded by the Marie Curie Early Stage Training grant (MEST-CT-2004-8201). The authors would like to thank all of our participants, and thank Marco Tiemann, Dragan Sekulovski, William Green, Jan Engel, Armin Kohlrausch, Sam Gosling, Peter Jason Rentfrow, and our anonymous reviewers for their help or advice toward producing this work.



Dunn et al. 17

NotesParticipants’ music listening behavior for song count and listening duration included all data from songs selected multiple times.There was missing data for musical training, resulting in a smaller df in the denominator than expected.

ReferencesAll Music Guide (AMG). (2008). All music guide. Retrieved 17 April 2008 from http://www.allmusic.com Arnett, J. (1992). The soundtrack of recklessness: Musical preferences and reckless behavior among ado-

lescents. Journal of Adolescent Research, 7(3), 313–331.Aucouturier, J. -J., & Pachet, F. (2003). Representing musical genre: A state of the art. Journal of New Music

Research, 32(1), 83–93.British Phonographic Industry (BPI). (2008). Sales by type of music, 2007. Retrieved 10 December 2008

from http://www.bpi.co.uk Cattell, R. B., & Anderson, J. C. (1953). The measurement of personality and behavior disorders by the

I.P.A.T. music preferences test. The Journal of Applied Psychology, 37(6), 446–454.Chamorro-Premuzic, T., & Furnham, A. (2007). Personality and music: Can traits explain how people use

music in everyday life? British Journal of Psychology, 98(2), 175–185.Costa, P. T., & McCrae, R. R. (1992). NEO PI-R professional manual. Odessa, FL: Psychological Assessment

Resources.Delsing, M. J. M. H., Ter Bogt, T. F. M., Engels, R. C. M. E., & Meeus, W. H. J. (2008). Adolescents’ music

preferences and personality characteristics. European Journal of Personality, 22(2), 109–130.George, D., Stickle, K., Rachid, F., & Wopnford, A. (2007). The association between types of music enjoyed

and cognitive, behavioral, and personality factors of those who listen. Psychomusicology, 19(2), 32–56.Goldberg, L. R. (2006). Doing it all bass-ackwards: The development of hierarchical factor structures from

the top down. Journal of Research in Personality, 40, 347–358.Hoekstra, H. A., Ormel, J., & de Fruyt, F. (2003). NEO-PI-R/NEO-FFI big five persoonlijkheidsvragenlijst:

Handleiding [NEO-PI-R/NEO-FFI big five personality inventory: Professional manual]. Lisse: Harcourt Assessment.

Information Technology (IT) Facts. (2008). 2007 US music sales by genre. Retrieved 15 February 2008 from http://www.itfacts.biz/2007-us-music-sales-by-genre/9354

John, O. P., & Srivastava, S. (1999). The Big Five trait taxonomy: History, measurement, and theoretical perspectives. In L. A. Pervin & O. P. John (Eds.), Handbook of personality: Theory and research (2nd ed., pp. 102–138). New York: Guilford Press.

Jöreskog, K., & Sörbom, D. (2007). LISREL 8.80. Scientific Software International, Inc.Levitin, D. J. (2006). This is your brain on music: The science of a human obsession. New York: Dutton/Penguin.Litle, P., & Zuckerman, M. (1986). Sensation seeking and music preferences. Personality and Individual

Differences, 7(4), 575–577.McNamara, L., & Ballard, M. E. (1999). Resting arousal, sensation seeking, and music preference. Genetic,

Social, and General Psychology Monographs, 125(3), 229–250.North, A. C., & Hargreaves, D. J. (1999). Music and adolescent identity. Music Education Research, 1(1),

75–92.O’Connor, B. P. (2000). SPSS and SAS programs for determining the number of components using par-

allel analysis and Velicer’s MAP test. Behavior Research Methods, Instruments, & Computers, 32(3), 396–402.

Rawlings, D., Barrantes i Vidal, N., & Furnham, A. (2000). Personality and aesthetic preference in Spain and England: Two studies relating sensation seeking and openness to experience to liking for paintings and music. European Journal of Personality, 14(6), 553–576.

Rawlings, D., Hodge, M., Sherr, D., & Dempsey, A. (1995). Toughmindedness and preference for musical excerpts, categories and triads. Psychology of Music, 23(1), 63–80.




Rentfrow, P. J., & Gosling, S. D. (2003). The do re mi’s of everyday life: The structure and personality correlates of music preferences. Journal of Personality and Social Psychology, 84(6), 1236–1256.

Rentfrow, P. J., & Gosling, S. D. (2006). Message in a ballad: The role of music preferences in interpersonal perception. Psychological Science, 17(3), 236–242.

SPSS. (2006). SPSS Statistics (Version 15.0): SPSS Inc.Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston, MA: Allyn and

Bacon.Zweigenhaft, R. L. (2008). A do re mi encore: A closer look at the personality correlates of music prefer-

ences. Journal of Individual Differences, 29(1), 45–55.

Peter Dunn has a background in Experimental Psychology and Human–Computer Interaction. Originally from Ottawa, Canada, he received his PhD from the University of Technology in Eindhoven, The Netherlands (TU/e) in January 2010. He is currently a User Experience Consultant at Flow Interactive in London, England. His research interests are focused on user/personality modelling, psychometrics, and user-centred design.

Boris de Ruyter, after his graduation, worked as a research assistant in experimental psychol-ogy at the University of Antwerp. Since 1994 he has been with Philips Research, where he is appointed as principal scientist. His research focuses on user modelling and psychometrics. Since 1999 he has been leading a research team of behavioral scientists that contributes to the Lifestyle research program of Philips Research. He is an author of multiple international pub-lications and owns numerous patents.

Don G. Bouwhuis has a background in cognitive science, mathematics, and computer science. After a career in industrial research he became Professor of Cognitive Ergonomics at the University of Technology, Eindhoven NL. His current research interests are centered on evalu-ation metrics for usability and user experience, on telecare and on the perceptual experience of novel media, like 3D TV and multimedia recommender systems.



Psychology of Music - Jurusan Informatikafrdaus/PenelusuranInformasi/... · 4 Psychology of Music Despite a concerted effort since 2003 to measure personality in terms of the Big

Documents